Identifying Organizational Effects on Medical Productivity Robert S. Huckman AcademyHealth

advertisement
Identifying Organizational Effects on
Medical Productivity
Robert S. Huckman
AcademyHealth
June 29, 2009
Preliminaries
• Medical productivity = quality-adjusted output per unit of
input
– Not just “how much” but also “how well”
• Assumption: Hospitals are points of interaction between
professionals
• Wide variation in the determinants of medical productivity
(Wennberg and Gittelsohn, 1973)
• Broad research question: To what extent do organizational
(i.e., hospital) factors impact medical productivity?
– Technological choice (what gets done)*
– Operational performance (how well things get done) for any given
technology**
* Robert S. Huckman, 2003. “The Utilization of Competing Technologies within the Firm: Evidence from Cardiac
Procedures,” Management Science, 49(5): 599-617.
** Robert S. Huckman and Gary P. Pisano, 2006. “The Firm Specificity of Individual Performance: Evidence from Cardiac
Surgery,” Management Science, 52(4): 473-488.
2
(Some) Organizational Determinants of
Medical Productivity
ENTRY INTO
SYSTEM
TECHNOLOGIAL
CHOICE
OPERATIONAL
PERFORMANCE
Decision to
Seek Care
Decision
Regarding
Treatment
Quality and
Cost of
Treatment
Diffuse turf
battles
Separate skill and
status
Leverage “residual
variation” from
multiple sources of
data
Build team
familiarity
Managerial
Objective
Separate
individual and
organizational
effects
Empirical
Challenge
Leverage
individual
mobility
Approach
3
Research Setting: Major Cardiac
Procedures
Source: Thomas Burton, “Bypass Surpasses Angioplasty in Study,” Wall Street Journal,
May 4, 2004
4
Comparison of Bypass Surgery and
Angioplasty
Angioplasty
Bypass Surgery
1970s
1960s
Cardiologist
Cardiac surgeon
Catheterization Lab
Cardiac OR
Severity of Patient
Condition
Low
High
Patient Discomfort
Low
High
Medium (with rapid
improvement)
High (with slight
improvement)
High
Very high
Date of Introduction
Type of Physician
Location (in Hospital)
Effectiveness
Financial Return for
Hospitals
5
Bypass Surgery and Angioplasty in
New York State
Procedure Rate/1,000 Population Age 45+
6.00
5.00
4.00
3.00
2.00
1.00
1982
1984
1986
1988
1990
1992
1994
1996
1998
2000
2002
Year
Bypass Surgery
Angioplasty
Source: David M. Cutler and Robert S. Huckman (2003), “Technological Development and Medical Productivity:
The Diffusion of Angioplasty in New York State," Journal of Health Economics.
6
Across-Hospital Variation in Bypass Rates
Adjusted Bypass Rate =
Bypass/(Bypass+Angioplasty)
Hospitals
Mean
Standard
Deviation
25th
Percentile
75th
Percentile
Low
29
0.048
0.032
0.030
0.063
Medium
29
0.359
0.107
0.274
0.434
High
29
0.847
0.043
0.815
0.869
Patient Severity
Note: The means and standard deviations are weighted by the total number of
revascularization cases at the given severity level for the three-year period from
1993 to 1995.
Source: Robert S. Huckman (2003), “The Utilization of Competing Technologies within the Firm: Evidence from
Cardiac Procedures," Management Science.
7
Possible Explanations for Variation
Across Hospitals
Differences in
• Patient characteristics
• Objective quality of bypass surgery relative to
angioplasty
• Relative profitability of bypass surgery and
angioplasty
• Overall financial performance
• Intensity of turf battles between surgeons and
cardiologists
8
Cardiac Turf Battles
• “Surgeons are not the best friends of angioplasts”*
• “…where the open-heart alternative is available, the number of
angioplasty procedures performed is usually small, and the
hospitals make no effort to recruit (physicians) who are skilled
in angioplasty.”*
• “‘It’s like the Wild West, the turf wars with the vascular
surgeons…You’ve got interventional cardiologists,
interventional neuroradiologists, interventional neurologists all
wanting in.’” (Alejandro Bernstein, M.D., neuroradiologist at
Beth Israel Medical Center, New York)**
Sources: *Robert Hamilton, “Hospital Proposes Bypass Alternative,” New York Times, March 16, 1986;
**Joanne Kaufman, “The Uncut Version,” New York Magazine, October 13, 2003.
9
Data
• Patient-level records for every individual
receiving either bypass surgery or angioplasty in
New York, 1993-95
– ~30 hospitals and ~100,000 procedures in sample
• Substantial detail
– Characteristics of patient (e.g., age, gender, clinical
characteristics)
– Surgeon and hospital identifiers
– In-hospital mortality outcomes
• Data are used as the basis of public report cards
about bypass and angioplasty quality
10
Variables
• Dependent variable: Risk-adjusted rate of bypass relative
to angioplasty for a given hospital-year
• Key independent variables
– Cardiac surgeons’ status = hospital’s count of bypass citations in
the medical literature (1973-1992)
– Cardiologists’ status = hospital’s count of angioplasty citations in
the medical literature (1973-1992)
• Controls
– Objective quality for each procedure
 Bypass surgery: Risk-adjusted mortality
 Angioplasty: Risk-adjusted rate of emergency bypass following
angioplasty
– Hospital characteristics (e.g., size, lagged profitability, payer mix,
teaching status, county population density)
11
Measuring the Impact of Turf Battles
on Technological Choice
Dependent Variable: Risk-Adjusted Bypass Rate
Low
Severity
Average Value of
Dependent Variable
4.8%
Impact of 1 SD Increase in
Bypass Citations
1.2% *
Impact of 1 SD Increase in
Angioplasty Citations
0.5%
Medium
Severity
High
Severity
36.0%
84.7%
5.2% ***
-0.8%
2.4% ***
-0.9% **
Emergency
AMI
93.4%
2.1% ***
0.2%
*,**,*** denote statistical significance at the 10%, 5%, and 1% levels, respectively
Source: Huckman (2003)
12
ENTRY INTO
SYSTEM
TECHNOLOGIAL
CHOICE
OPERATIONAL
PERFORMANCE
Decision to
Seek Care
Decision
Regarding
Treatment
Quality and
Cost of
Treatment
Diffuse turf
battles
Separate skill and
status
Leverage “residual
variation” from
multiple sources of
data
Build team
familiarity
Managerial
Objective
Separate
individual and
organizational
effects
Empirical
Challenge
Leverage
individual
mobility
Approach
13
Team Familiarity and Surgical
Outcomes
• Many physicians in the United States are freelancers
 may work at multiple hospitals
• Hospitals employ other team members (e.g., nurses,
techs, anesthesiologists) who may be important in
determining performance
Test for importance of familiarity:
Can the experience of a surgeon be fully transported
across hospitals?
14
Are Volume-Outcome Effects For
Individual Surgeons Hospital Specific?
Risk-Adjusted Mortality
Rate for Surgeon S at
Hospital A
Hospital B?
Hospital B?
Hospital A Hospital B?
Surgeon S Volume
15
Why Might Hospital-Specific Volume
Matter?
• Familiarity: high volume is associated with the productive
benefits of organizational familiarity (e.g., team stability,
experience with firm-specific processes)
• Status/Influence: high volume is associated with the
political benefits of access to more or superior resources
from the organization (e.g., staff, financial support)
• Adding proxies for status/influence does not affect base
findings
– Surgeon’s share of hospital’s bypass volume in prior quarter
(financial importance)
– Surgeon’s share of hospital’s total citations for bypass related
publications (status)
16
Research Setting
• Data: Patient-level records for every bypass
surgery in Pennsylvania, 1994-95
– ~34,000 patients, 40 hospitals, and 190 surgeons
• Detail is similar to that in New York database and
is also used for public report cards
• Why Pennsylvania?
– High degree of splitting activity (~25% of surgeons)
– Splitting occurs roughly simultaneously—reduces
concerns about changes in surgeon performance over
time
17
Empirical Specification
• Logistic regression using patient-level observations
(surgeon s and hospital h)
• Dependent variable: Indicator equal to one if patient died in the
hospital; zero otherwise
• Key independent variables
– Surgeon s volume of cases at hospital h in prior quarter
– Surgeon s volume of cases at hospitals other than h in prior quarter
• Controls
– Surgeon quality: Surgeon s risk-adjusted mortality rate (across all
hospitals) in prior quarter
– Hospital quality: Hospital h risk-adjusted mortality rate (across all
surgeons) in prior quarter
– Patient characteristics (demographic and clinical)
– Influence proxies (robustness)
18
Results
Did Bypass Patient
Die in Hospital?
Impact of One Standard Deviation Increase in:
HOSPCASE : Surgeon s Cases at Hospital h (Prior
Quarter)
-0.32% ***
OTHCASE : Surgeon s Cases at Hospitals Other
Than h (Prior Quarter)
-0.01%
Average Predicted Probability of Mortality (Evaluated at
Means of Independent Variables)
Level at which HOSPCASE is significantly different from
OTHCASE
1.76%
1%
Note: *** denotes statistical significance at the 1% level.
Source: Huckman and Pisano (2006)
19
Summing Up
Broad Conclusion
Managerial
Objective
Methodological
Challenge
Methodological
“Solution”
Organizations may not
make decisions
concerning
technological choice,
but they do affect the
way in which such
decisions are made
Mitigate
negative effects
of turf battles
between groups
Identifying the
intensity of turf
battles identifying
status independent of
skill
Leverage “residual
variation” by using
multiple sources of data
The performance of
highly-skilled workers
is not perfectly
transferrable across
firms  organizations
matter
Build team
familiarity
Separating individual
and organizational
effects
Leverage individual
mobility
20
Download